[PATCH v4] Do not decode url protocol.

Subsystems: the rest

DORMANTno replies

2 messages, 2 authors, 2016-06-15 · open the first message on its own page

[PATCH v4] Do not decode url protocol.

From: Pascal Obry <hidden>
Date: 2016-06-15 22:48:59

When using the protocol git+ssh:// for example we do not want to
decode the '+' as a space. The url decoding must take place only
for the server name and parameters.

This fixes a regression introduced in 9d2e942.
---
 url.c |   19 +++++++++++++++----
 1 files changed, 15 insertions(+), 4 deletions(-)

Ok, so this is the fourth version of this patch. Thanks again Matthieu
for the review. I think this time I got the place for the message right :)

Anyway, I think this time we properly skip the protocol decoding when
needed.
diff --git a/url.c b/url.c
index cd32b92..0eb7fb3 100644
--- a/url.c
+++ b/url.c
@@ -67,12 +67,23 @@ static int url_decode_char(const char *q)
        return val;
 }

-static char *url_decode_internal(const char **query, const char *stop_at)
+static char *url_decode_internal(const char **query, const char *stop_at,
+               int with_protocol)
 {
        const char *q = *query;
+       const char *first_slash;
        struct strbuf out;

        strbuf_init(&out, 16);
+
+       /* Skip protocol if present. */
+       if (with_protocol) {
+         first_slash = strchr(*query, '/');
+
+         while (q < first_slash)
+               strbuf_addch(&out, *q++);
+       }
+
        do {
                unsigned char c = *q;
@@ -104,15 +115,15 @@ static char *url_decode_internal(const char
**query, const char *stop_at)

 char *url_decode(const char *url)
 {
-       return url_decode_internal(&url, NULL);
+       return url_decode_internal(&url, NULL, 1);
 }

 char *url_decode_parameter_name(const char **query)
 {
-       return url_decode_internal(query, "&=");
+       return url_decode_internal(query, "&=", 0);
 }

 char *url_decode_parameter_value(const char **query)
 {
-       return url_decode_internal(query, "&");
+       return url_decode_internal(query, "&", 0);
 }
-- 
1.7.1.426.gb436.dirty

-- 

--|------------------------------------------------------
--| Pascal Obry                           Team-Ada Member
--| 45, rue Gabriel Peri - 78114 Magny Les Hameaux FRANCE
--|------------------------------------------------------
--|    http://www.obry.net  -  http://v2p.fr.eu.org
--| "The best way to travel is by means of imagination"
--|
--| gpg --keyserver keys.gnupg.net --recv-key F949BD3B

Re: [PATCH v4] Do not decode url protocol.

From: Jeff King <hidden>
Date: 2016-06-15 22:48:59

On Tue, Jun 22, 2010 at 10:16:57PM +0200, Pascal Obry wrote:
When using the protocol git+ssh:// for example we do not want to
decode the '+' as a space. The url decoding must take place only
for the server name and parameters.

This fixes a regression introduced in 9d2e942.
Sorry, this is completely my fault. I obviously didn't test with
git+ssh. I looked at adding a new test case, but I don't think there's
an easy way. The only way to trigger it is by using "git+ssh" or
"ssh+git"; any other protocol (real or fake) will be handled by a custom
protocol handler and won't execute this broken code at all.

The patch looks reasonable (and I agree with all of Matthieu's comments
thus far). With respect to the "what if there is no slash" issue
discussed earlier, I prefer to be a bit defensive about possible inputs
in library-ish functions like this. But:
+
+       /* Skip protocol if present. */
+       if (with_protocol) {
+         first_slash = strchr(*query, '/');
+
+         while (q < first_slash)
+               strbuf_addch(&out, *q++);
+       }
+
I think this actually works OK, given that "q < first_slash" will be
false if first_slash is NULL, assuming NULL is all-bytes-zero or some
low number (which is not guaranteed by the standard, but is a pretty
practical assumption these days).

However, you could make things a little more efficient by handing the
whole thing to strbuf at once:

  if (with_protocol) {
          const char *first_slash = strchr(q, '/');
          if (first_slash) {
                  strbuf_add(&out, q, first_slash - q);
                  q = first_slash;
          }
  }

which of course needs first_slash to be checked explicitly.  Not that
the efficiency probably matters in practice.

-Peff
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help